Public concern has grown about the safety of the U.S. pharmaceutical supply due to well-publicized drug withdrawals or adverse reactions. The Institute of Medicine recently conducted a study of the oversight options available to the Food and Drug Administration (FDA) and concluded that substantial improvements are required in the way that post-marketing drug safety issues are identified and handled. The most restrictive labeling tool used by the FDA is mandating a boxed warning. However, boxed warnings are issued in the midst of a complex information environment that includes direct to consumer advertising (DTCA), general news media coverage, scientific journal publications, and direct to physician marketing. No research has been conducted to determine how the alternative sources of information on safety work simultaneously to prompt changes in prescribing behavior and whether the boxed warning mechanism has any marginal effect on pharmaceutical use. This is particularly important since the FDA is poised to devote substantial resources and time into revising its safety system, which would be inefficient if the boxed warning tool has little or no marginal effect on prescribing behaviors. This project will therefore have three aims: (1) examine the effects of public information - including media reports, DTCA, physician-directed marketing, and clinical research - and black box warnings on the flow of patients into physician practices;(2) examine the effects of public information on the likelihood of appropriate drug prescribing;and (3) examine the impact of FDA issued boxed warnings on the likelihood of appropriate prescribing for specific medications. This proposal represents a competing continuation of a completed research grant funded by the Agency for Healthcare Research and Quality titled "Impact of Direct to Consumer Pharmaceutical Marketing" (1R01HS011326-01A2). The specific aims will be achieved by estimating empirical models that examine the how prescribing is affected by public information and the issuance of advisories and boxed warnings by the FDA. We will use a unique electronic medical record data base of over 8 million patients in 119 geographically dispersed physician practices. We will do so in the context of the following three disease / drug classes: 1) long-acting [unreadable]2-agonists for asthma management, 2) antidepressant use in adolescents, and 3) use of selective and nonselective NSAIDs for pain management.